A scene display system and method based on 3D virtual simulation

By designing a 3D virtual simulation scene display system, using depth cameras, sensors and WebGL technology, the problem of the virtual simulation scene display system being unable to be customized and visualized is poor, and efficient scene simulation and data analysis are achieved.

CN117830575BActive Publication Date: 2024-11-12GUOWENJU (BEIJING) CULTURE TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410021888.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-01-08
Publication Date
2024-11-12
Estimated Expiration
2044-01-08

AI Technical Summary

Technical Problem

The existing virtual simulation scenario display system cannot be customized according to specific needs, and the visual expression effect of data results is not intuitive, so complex scenarios cannot be simulated.

Method used

A scene display system based on 3D virtual simulation was designed, including data collection, processing, simulation simulation, topological structure generation and display subsystem, through depth cameras, sensors, TCP/IP protocol, TCP/IP protocol, WebGL simulation technology, etc. Components, obtain and process scene data, generate three-dimensional simulation models, and build a tree-like structure for display.

Benefits of technology

It realizes a more experiential and interactive environment, can customize simulation scenarios according to specific needs, provide more intuitive visual expression effects, and supports real-time data collection and analysis.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117830575B_ABST
    Figure CN117830575B_ABST
Patent Text Reader

Abstract

The present invention discloses a scene display system and method based on 3D virtual simulation, and the technical field belongs to the field of virtual simulation, including: obtaining scene data information based on a depth camera and a sensor; processing the scene data information to obtain a label data set; performing a three-dimensional simulation on the label data set through a three-dimensional simulation software to obtain a three-dimensional simulation model, adding a map texture material to the three-dimensional simulation model and partitioning it to generate a scene partition simulation model; calculating the correlation of the scene partition simulation model, constructing a tree structure of the scene partition simulation model based on the correlation, and generating a scene tree model; and displaying the scene tree model in a browser based on WebGL simulation technology. The 3D virtual simulation system of the topological structure of the present invention not only provides a more experiential and interactive environment, but also can simulate a variety of scenes in a more economical and efficient way.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the field of virtual simulation, and in particular relates to a scene display system and method based on 3D virtual simulation. Background Art

[0002] Virtual simulation, also known as virtual reality technology or simulation technology, is a technology that uses a virtual system to imitate another real system. The artificial environment of three-dimensional information composed of computer hardware, software and various sensors - the virtual environment, realistically simulates the objects and environment of the real world. Virtual simulation technology is a product of combining simulation technology with virtual reality technology on the basis of the rapid development of information technology such as multimedia technology, virtual reality technology and network communication technology. It is a more advanced simulation technology. Virtual simulation technology is characterized by building a complete virtual environment that is unified for the entire system, and integrating and controlling a large number of entities through the virtual environment. The entity can be a simulator, or other virtual simulation system, or it can be represented by some simple mathematical models. Entities interact with each other in the virtual environment, or interact with the virtual environment to show the real characteristics of the objective world. The integrated, virtualized and networked characteristics of virtual simulation technology fully meet the development needs of modern simulation technology.

[0003] However, the current virtual simulation scene display system has the following shortcomings: 1. The current virtual simulation scene cannot be customized according to specific needs; 2. Although the simulation results obtained by semi-physical simulation are more realistic than those obtained by full digital simulation, its data results are also limited to numbers and curves, lacking a more intuitive visual expression effect, and cannot simulate the scene. Summary of the invention

[0004] The purpose of the present invention is to provide a scene display system and method based on 3D virtual simulation to solve the problems existing in the above-mentioned prior art.

[0005] To achieve the above object, the present invention provides a scene display system based on 3D virtual simulation, comprising:

[0006] A data collection subsystem, used to obtain scene data information based on depth cameras and sensors;

[0007] A processing subsystem is connected to the data collection subsystem via the TCP / IP protocol, and is used to process the scene data information, perform data feature engineering on the processed scene data information based on the device type, and then generate labels to obtain a label data set;

[0008] A simulation subsystem is connected to the processing subsystem via the TCP / IP protocol, and is used to perform a three-dimensional simulation on the label data set through a three-dimensional simulation software to obtain a three-dimensional simulation model, add a map texture material to the three-dimensional simulation model and partition it, and generate a scene partition simulation model;

[0009] A topology structure generation subsystem is connected to the simulation subsystem via the TCP / IP protocol, and is used to calculate the correlation of the scene partition simulation model, construct a tree structure of the scene partition simulation model based on the correlation, and generate a scene tree model;

[0010] The display subsystem is connected to the topology structure generation module through the TCP / IP protocol and is used to display the scene tree model in a browser based on the WebGL simulation technology.

[0011] Preferably, the data collection subsystem includes:

[0012] Depth camera, used to shoot the scene and collect distance information to obtain scene point cloud information;

[0013] The sensor module is used to obtain the real description of the scene based on distributed sensors, annotate the identifiable targets, and generate scene annotation information;

[0014] An integration module is used to integrate the scene point cloud information and the scene annotation information to obtain the scene data information.

[0015] Preferably, the scene point cloud information collected by the depth camera includes spatial coordinate information, color information, intensity information, normal information, timestamp and classification information;

[0016] The scene annotation information collected by the sensor module includes audio data, motion data and position data.

[0017] Preferably, the processing subsystem includes:

[0018] A data cleaning module is used to clean the scene data information for missing values, fill in the missing values ​​through a regression analysis method, obtain a cleaned data set, and proofread the cleaned data set to obtain a processed data set;

[0019] A standardization module, used for performing one-hot encoding on the categorical features in the processed data set and standardizing the numerical features in the processed data set to obtain a feature data set;

[0020] A definition module defines the relevance labels of the feature data set based on the experience labels and the physical and electrical relationships between scene objects to generate the label data set.

[0021] Preferably, the simulation subsystem includes:

[0022] A rendering module, used for preparing vertex and fragment shaders, transferring the vertex data to the fragment shader for a rendering cycle, and obtaining a rendering model;

[0023] A framework import module is used to simplify the rendering process of the rendering model based on the WebGL library to obtain an initial simulation model;

[0024] An optimization module, used for optimizing the performance of the initial simulation model based on shader programs, batch rendering calls and texture compression technology to obtain the three-dimensional simulation model;

[0025] A texture management module, used for applying textures to the three-dimensional simulation model to generate a simulation model supporting texture mapping;

[0026] A partitioning module is used to partition the simulation model of the texture map based on labels to obtain the scene partition simulation model.

[0027] Preferably, the topology structure generation subsystem includes:

[0028] A perceptron generation module is used to train a multi-layer perceptron based on a historical scene relevance sample set to obtain a relevance calculation model;

[0029] A correlation generation module, used to calculate the correlation of the scene partition simulation model based on the correlation calculation model to obtain the block correlation;

[0030] A topology structure building module is used to connect the scene partition simulation model based on the block association degree and generate a tree structure to obtain the scene tree model.

[0031] Preferably, the sensor generation module includes:

[0032] A perceptron construction submodule, used to construct an initial multilayer perceptron, improve the initial multilayer perceptron, and generate the multilayer perceptron;

[0033] The training submodule is used to train the multi-layer perceptron based on the historical scene relevance sample set to obtain the relevance calculation model.

[0034] Preferably, the sensor construction submodule includes:

[0035] Network generation unit, building a deep residual network and constructing an initial multi-layer perceptron;

[0036] A residual block generation unit, which adds the outputs of the main path and the shortcut in the deep residual network to form a residual connection, and matches the dimensions of the residual connection through identity mapping to generate a residual block;

[0037] The construction unit adds the residual block between each hidden layer of the initial multilayer perceptron to generate the multilayer perceptron.

[0038] Preferably, the training submodule includes:

[0039] A division unit, used for dividing the historical scene relevance sample set into a training set and a verification set;

[0040] A perceptron training unit, configured to train the multilayer perceptron using the training set, select a cross entropy loss function, and minimize the cross entropy loss function by a stochastic gradient descent algorithm to obtain a trained perceptron;

[0041] A verification unit, used to verify the performance of the trained perceptron using the verification set to obtain a verification result;

[0042] An optimization unit is used to optimize the learning rate and the number of hidden layers of the training perceptron based on the verification result to obtain the association calculation model.

[0043] To achieve the above object, the present invention also provides a scene display method based on 3D virtual simulation, comprising:

[0044] Acquire scene data information based on depth cameras and sensors;

[0045] Processing the scene data information, performing data feature engineering on the processed scene data information based on the device type, and then generating labels to obtain a label data set;

[0046] Performing three-dimensional simulation on the label data set by using three-dimensional simulation software to obtain a three-dimensional simulation model, adding map texture materials to the three-dimensional simulation model and partitioning it to generate a scene partition simulation model;

[0047] Calculating the correlation of the scene partition simulation model, constructing a tree structure of the scene partition simulation model based on the correlation, and generating a scene tree model;

[0048] The scene tree model is displayed in a browser based on WebGL simulation technology.

[0049] The technical effects of the present invention are:

[0050] The 3D virtual simulation system of the topological structure of the present invention not only provides a more experiential and interactive environment, but also can simulate multiple scenarios in a more economical and efficient way. At the same time, the system of the topological structure usually has data collection and analysis functions, which can collect and analyze data in the simulation process in real time to help users understand the situation and make corresponding adjustments. The system of the topological structure can be customized according to specific needs, arranged and adjusted according to the scene requirements to meet the user's specific simulation needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0052] Figure 1 Schematic diagram of a scene display system based on 3D virtual simulation in an embodiment of the present invention;

[0053] Figure 2 This is a flow chart of a scene display method based on 3D virtual simulation in an embodiment of the present invention. DETAILED DESCRIPTION

[0054] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0055] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0056] Embodiment 1

[0057] like Figure 1 As shown, this embodiment provides a scene display system based on 3D virtual simulation, including:

[0058] A data collection subsystem, used to obtain scene data information based on depth cameras and sensors;

[0059] A processing subsystem is connected to the data collection subsystem via the TCP / IP protocol, and is used to process the scene data information, perform data feature engineering on the processed scene data information based on the device type, and then generate labels to obtain a label data set;

[0060] A simulation subsystem is connected to the processing subsystem via the TCP / IP protocol, and is used to perform a three-dimensional simulation on the label data set through a three-dimensional simulation software to obtain a three-dimensional simulation model, add a map texture material to the three-dimensional simulation model and partition it, and generate a scene partition simulation model;

[0061] A topology structure generation subsystem is connected to the simulation subsystem via the TCP / IP protocol, and is used to calculate the correlation of the scene partition simulation model, construct a tree structure of the scene partition simulation model based on the correlation, and generate a scene tree model;

[0062] The display subsystem is connected to the topology structure generation module through the TCP / IP protocol and is used to display the scene tree model in a browser based on the WebGL simulation technology.

[0063] The WebGL simulation technical parameter requirements in this embodiment include: (1) Browser compatibility: support Mozilla Firefox 52+, Google Chrome 57+, Apple Safari 11+, Microsoft Edge 16+.

[0064] (2) Support Web audio: handle audio playback and mixing.

[0065] (3) Supports hardware accelerated rendering using GPU.

[0066] (4) Support diffuse reflection, specular reflection, ambient lighting, point light source, and parallel light source.

[0067] (5) The surface of an object is uneven compared to light. From a macroscopic perspective, when a beam of light shines on the surface of an object, for ideal diffuse reflection, the surface has irregular and randomly distributed uneven reflective surfaces, and the reflection of light is non-directional. The light reflected from the object into the eye has nothing to do with the observation position of the scene camera.

[0068] (6) Texture mapping support: Display images in png, jpg and other formats in the WebGL 3D scene. By adjusting the value of the transparency A component, color overlay can be achieved to simulate transparent or translucent glass effects.

[0069] (7) The supported 3D model formats include .fbx, .dae (Collada), .3ds, .dxf, .max, .c4d, .blender, .ma / mb and .obj files.

[0070] (8) Support switching shader programs: model materials and textures can be changed.

[0071] (9) Supports WebGL transparency and alpha fusion, and can simulate translucent effects through color fusion technology.

[0072] (10) The platform is based on the Three.js WebGL framework and supports rendering of simulated three-dimensional scenes through renderers and cameras.

[0073] (11) Supports Mesh objects, Light objects, Group objects, 3D objects, and Camera objects to build Scene objects. Scene objects define the most basic elements of a scene and include properties such as background color and fog. These objects display their respective positions and directions through a tree structure with a clear hierarchical relationship. The position and direction of a child object are always relative to the parent object.

[0074] (12) The scene supports placing multiple cameras in one scene, supports multiple materials, supports textures, supports multiple geometric bodies such as spheres, cubes, planes, etc. Supports different light source objects.

[0075] Further optimizing the scheme, the data collection subsystem includes:

[0076] Depth camera, used to shoot the scene and collect distance information to obtain scene point cloud information;

[0077] The data acquired by the depth camera is a set of three-dimensional points, and the coordinates of each point represent the position of an object or surface in the scene. The depth camera measures the distance from the object to the camera and can convert this distance information into a point cloud in three-dimensional space. Point clouds are often used in applications such as three-dimensional modeling, virtual reality, augmented reality, and robotic perception.

[0078] The sensor module is used to obtain a true description of the scene based on distributed sensors, and to mark identifiable targets to generate scene annotation information; the data used by the sensor to obtain a true description of the scene may include a variety of information.

[0079] An integration module is used to integrate the scene point cloud information and the scene annotation information to obtain the scene data information.

[0080] Further optimizing the scheme, the scene point cloud information collected by the depth camera includes spatial coordinate information, color information, intensity information, normal information, timestamp and classification information;

[0081] The scene annotation information collected by the sensor module includes audio data, motion data and position data.

[0082] Further optimizing the solution, the processing subsystem includes:

[0083] A data cleaning module is used to clean the scene data information for missing values, fill in the missing values ​​through a regression analysis method, obtain a cleaned data set, and proofread the cleaned data set to obtain a processed data set;

[0084] A standardization module, used for performing one-hot encoding on the categorical features in the processed data set and standardizing the numerical features in the processed data set to obtain a feature data set;

[0085] The processing of standardized modules is as follows:

[0086] Loading data:

[0087] Use Pandas or other data processing libraries to load the data set of this embodiment.

[0088] Use OneHotEncoder for one-hot encoding:

[0089] Create a OneHotEncoder object and use it to convert the selected categorical features.

[0090] The parameter sparse is set to False to indicate that the output is a dense matrix, and drop = 'first' means to delete the first encoding of each feature to avoid multicollinearity.

[0091] Merge the one-hot encoded result with the original data:

[0092] Merge the result of one-hot encoding with the original data by concatenating its columns into the original DataFrame.

[0093] Complete one-hot encoding:

[0094] Based on the experience labels and the physical and electrical relationships between the devices, the relevance labels of the feature data set are defined to generate the label data set. Relevance labels are defined for the samples in the training set. These labels can be derived from the experience of professionals or inferred based on the physical and electrical relationships between the devices.

[0095] A definition module defines the relevance labels of the feature data set based on the experience labels and the physical and electrical relationships between scene objects to generate the label data set.

[0096] Further optimizing the scheme, the simulation subsystem includes:

[0097] A rendering module, used for preparing vertex and fragment shaders, transferring the vertex data to the fragment shader for a rendering cycle, and obtaining a rendering model;

[0098] Specifically:

[0099] Writing vertex and fragment shaders: Create vertex shaders and fragment shaders to handle rendering of vertex data and fragment data.

[0100] Prepare vertex data: define the vertices and their attributes (such as position, color, normal, etc.) and transfer them to the shader.

[0101] Writing a rendering loop: Create a rendering loop, which includes steps such as clearing the canvas, setting up the view and projection matrices, preparing shader programs, passing vertex data, etc.

[0102] A framework import module is used to simplify the rendering process of the rendering model based on the WebGL library to obtain an initial simulation model;

[0103] Specifically:

[0104] Simplify the 3D rendering process by using existing WebGL libraries or frameworks such as Three.js, Babylon.js, etc. These libraries provide high-level abstractions and functions that make it more convenient to create and manage 3D scenes.

[0105] An optimization module, used for optimizing the performance of the initial simulation model based on shader programs, batch rendering calls and texture compression technology to obtain the three-dimensional simulation model;

[0106] The texture management module is used to apply textures to the three-dimensional simulation model to generate a simulation model that supports texture mapping; load and use external three-dimensional model files (such as OBJ, GLTF and other formats) and texture images, and apply them to objects in the scene.

[0107] Also use shader programs, batch render calls, use texture compression and other techniques to improve performance.

[0108] The source code of the shader is:

[0109] / / Get WebGL context const canvas = document.getElementById('myCanvas'); const gl = canvas.getContext('webgl'); / / Vertex shader source code const vertexShader Source = ` / / attribute is used to receive vertex data attributevec3aVertexPosition; void main(){ / / pass vertex position to fragment shader gl_Position = vec4(aVertexPosition, 1.0);}`; / / fragment shader source code const fragmentShaderSource = `void main(){ / / set fragment color to red gl_FragColor = vec4(1.0, 0.0, 0.0, 1.0);}`; / / create vertex shader const vertexShader = gl.createShader(gl.VERTEX_SHADER); gl.shaderSource(vertexShader, vertexShaderSource); gl.compileShader(vertexShader); / / create fragment shader const fragmentShader = gl.createShader(gl.FRAGMENT_SHADER); gl.shaderSource(fragmentShader, fragmentShaderSource); gl.compileShader(fragmentShader); / / create shader program const shaderProgram = gl.createProgram gram(); gl.attachShader(shaderProgram, vertexShader); gl.attachShader(shaderProgram, fragmentShader); gl.linkProgram(shaderProgram); / / Use the shader program gl.useProgram(shaderProgram);

[0110] Consider mobile devices: Ensure good performance on mobile devices and avoid excessive calculations and resource usage.

[0111] A partitioning module is used to partition the simulation model of the texture map based on labels to obtain the scene partition simulation model.

[0112] Further optimizing the scheme, the topology structure generation subsystem includes:

[0113] A perceptron generation module is used to train a multi-layer perceptron based on a historical scene relevance sample set to obtain a relevance calculation model;

[0114] A correlation generation module, used to calculate the correlation of the scene partition simulation model based on the correlation calculation model to obtain the block correlation;

[0115] A topology structure building module is used to connect the scene partition simulation model based on the block association degree and generate a tree structure to obtain the scene tree model.

[0116] Technical parameter requirements for model making (1) Model unit and position The three-dimensional model uses centimeters (cm) as the unit. The center point of all model groups is the origin (0, 0). The object model in the group is centered on itself without any displacement or rotation parameters.

[0117] (2) Model making standards

[0118] All models are made using the Polygon method, not the Nurbs modeling method.

[0119] The normals of the model are all pointing outwards, towards the correct side.

[0120] Under the premise of ensuring the effect, it can fully reflect the appearance of the 3D model, reasonably control the accuracy, reduce the number of model faces (less than 5000 faces) and the number of materials, and achieve the streamlining of data volume.

[0121] All models are exported in fbx and obj file formats. When exported in fbx file format, the textures are attached in media form, and a single exported file is no larger than 500MB.

[0122] All models have no missing faces, no flashing or overlapping faces, no deformed concave surfaces, and no mold penetration issues.

[0123] (3) Material Mapping Standards

[0124] The model uses Standard material, and texture materials can be added as needed.

[0125] The texture format is png file format, the color image is true color (24-bit color), and the image grayscale is not less than 256. The unit size is 2N power, the maximum size is not more than 1024*1024, the minimum size is not more than 16*16, and the resolution is not less than 72dpi.

[0126] In Photoshop image processing software, functions such as "transform", "free transform" and "rotate" are used to correct various textures, so that the corrected textures have a normal effect.

[0127] Further optimizing the scheme, the sensor generation module includes:

[0128] A perceptron construction submodule, used to construct an initial multilayer perceptron, improve the initial multilayer perceptron, and generate the multilayer perceptron;

[0129] The training submodule is used to train the multi-layer perceptron based on the historical scene relevance sample set to obtain the relevance calculation model.

[0130] Further optimizing the scheme, the perceptron construction submodule includes:

[0131] Network generation unit, building a deep residual network and constructing an initial multi-layer perceptron;

[0132] A residual block generation unit, which adds the outputs of the main path and the shortcut in the deep residual network to form a residual connection, and matches the dimensions of the residual connection through identity mapping to generate a residual block;

[0133] The residual block consists of two main branches, one is the main path (identity mapping) and the other is the shortcut.

[0134] The main path consists of a series of convolutional layers, while the shortcut adds the input directly to the main path output.

[0135] The design of the residual block can refer to the basic block structure in ResNet, such as two 3x3 convolutional layers and a BatchNormalization layer.

[0136] In the main path, a suitable activation function, such as ReLU, can be used to introduce non-linearity.

[0137] The output of the shortcut is added to the output of the main path to form a residual connection.

[0138] Here we need to ensure that the dimensions of the two match, which we achieve by using an identity mapping.

[0139] Add a residual block between each hidden layer of the multilayer perceptron.

[0140] The construction unit adds the residual block between each hidden layer of the initial multilayer perceptron to generate the multilayer perceptron.

[0141] Further optimizing the scheme, the training submodule includes:

[0142] A division unit, used for dividing the historical scene relevance sample set into a training set and a verification set;

[0143] A perceptron training unit, configured to train the multilayer perceptron using the training set, select a cross entropy loss function, and minimize the cross entropy loss function by a stochastic gradient descent algorithm to obtain a trained perceptron;

[0144] A verification unit, used to verify the performance of the trained perceptron using the verification set to obtain a verification result;

[0145] An optimization unit is used to optimize the learning rate and the number of hidden layers of the training perceptron based on the verification result to obtain the association calculation model.

[0146] Embodiment 2

[0147] like Figure 2 As shown, this embodiment provides a scene display method based on 3D virtual simulation, including:

[0148] Acquire scene data information based on depth cameras and sensors;

[0149] Processing the scene data information, performing data feature engineering on the processed scene data information based on the device type, and then generating labels to obtain a label data set;

[0150] Performing three-dimensional simulation on the label data set by using three-dimensional simulation software to obtain a three-dimensional simulation model, adding map texture materials to the three-dimensional simulation model and partitioning it to generate a scene partition simulation model;

[0151] Calculating the correlation of the scene partition simulation model, constructing a tree structure of the scene partition simulation model based on the correlation, and generating a scene tree model;

[0152] The scene tree model is displayed in a browser based on WebGL simulation technology.

[0153] The above is only a preferred specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A scene display system based on 3D virtual simulation, characterized in that: include: A data collection subsystem, used to obtain scene data information based on depth cameras and sensors; The data collection subsystem includes: Depth camera, used to shoot the scene and collect distance information to obtain scene point cloud information; The sensor module is used to obtain the real description of the scene based on distributed sensors, annotate the identifiable targets, and generate scene annotation information; An integration module, used to integrate the scene point cloud information and the scene annotation information to obtain the scene data information; The scene point cloud information collected by the depth camera includes spatial coordinate information, color information, intensity information, normal information, timestamp and classification information; The scene annotation information collected by the sensor module includes audio data, motion data and location data; A processing subsystem is connected to the data collection subsystem via the TCP / IP protocol, and is used to process the scene data information, perform data feature engineering on the processed scene data information based on the device type, and then generate labels to obtain a label data set; A simulation subsystem is connected to the processing subsystem via the TCP / IP protocol, and is used to perform a three-dimensional simulation on the label data set through a three-dimensional simulation software to obtain a three-dimensional simulation model, add a map texture material to the three-dimensional simulation model and partition it, and generate a scene partition simulation model; A topology structure generation subsystem is connected to the simulation subsystem via the TCP / IP protocol, and is used to calculate the correlation of the scene partition simulation model, construct a tree structure of the scene partition simulation model based on the correlation, and generate a scene tree model; The topology structure generation subsystem includes: A perceptron generation module is used to train a multi-layer perceptron based on a historical scene relevance sample set to obtain a relevance calculation model; A correlation generation module, used to calculate the correlation of the scene partition simulation model based on the correlation calculation model to obtain the block correlation; A topology structure building module, used to connect the scene partition simulation model based on the block association degree and generate a tree structure to obtain the scene tree model; The sensor generation module includes: A perceptron construction submodule, used to construct an initial multilayer perceptron, improve the initial multilayer perceptron, and generate the multilayer perceptron; A training submodule, used for training the multilayer perceptron based on a historical scene relevance sample set to obtain the relevance calculation model; The sensor construction submodule includes: Network generation unit, building a deep residual network and constructing an initial multi-layer perceptron; A residual block generation unit, which adds the outputs of the main path and the shortcut in the deep residual network to form a residual connection, and matches the dimensions of the residual connection through identity mapping to generate a residual block; A construction unit is configured to add the residual block between each hidden layer of the initial multilayer perceptron to generate the multilayer perceptron; The training submodule includes: A division unit, used for dividing the historical scene relevance sample set into a training set and a verification set; A perceptron training unit, configured to train the multilayer perceptron using the training set, select a cross entropy loss function, and minimize the cross entropy loss function by a stochastic gradient descent algorithm to obtain a trained perceptron; A verification unit, used to verify the performance of the trained perceptron using the verification set to obtain a verification result; An optimization unit, configured to optimize the learning rate and the number of hidden layers of the training perceptron based on the verification result to obtain the correlation calculation model; The display subsystem is connected to the topology structure generation module through the TCP / IP protocol and is used to display the scene tree model in a browser based on the WebGL simulation technology.

2. The scene display system based on 3D virtual simulation according to claim 1 is characterized in that: The processing subsystem includes: A data cleaning module is used to clean the scene data information for missing values, fill in the missing values ​​through a regression analysis method, obtain a cleaned data set, and proofread the cleaned data set to obtain a processed data set; A standardization module, used for performing one-hot encoding on the categorical features in the processed data set and standardizing the numerical features in the processed data set to obtain a feature data set; A definition module defines the relevance labels of the feature data set based on the experience labels and the physical and electrical relationships between scene objects to generate the label data set.

3. The scene display system based on 3D virtual simulation according to claim 1 is characterized in that: The simulation subsystem includes: A rendering module, used for preparing vertex and fragment shaders, transferring the vertex data to the fragment shader for a rendering cycle, and obtaining a rendering model; A framework import module is used to simplify the rendering process of the rendering model based on the WebGL library to obtain an initial simulation model; An optimization module, used for optimizing the performance of the initial simulation model based on shader programs, batch rendering calls and texture compression technology to obtain the three-dimensional simulation model; A texture management module, used for applying textures to the three-dimensional simulation model to generate a simulation model supporting texture mapping; A partitioning module is used to partition the simulation model of the texture map based on labels to obtain the scene partition simulation model.

4. The display method of the scene display system based on 3D virtual simulation according to claim 1, characterized in that: The following steps are involved: Acquire scene data information based on depth cameras and sensors; Processing the scene data information, performing data feature engineering on the processed scene data information based on the device type, and then generating labels to obtain a label data set; Performing three-dimensional simulation on the label data set by using three-dimensional simulation software to obtain a three-dimensional simulation model, adding map texture materials to the three-dimensional simulation model and partitioning it to generate a scene partition simulation model; Calculating the correlation of the scene partition simulation model, constructing a tree structure of the scene partition simulation model based on the correlation, and generating a scene tree model; The scene tree model is displayed in a browser based on WebGL simulation technology.